Two thermal evolution models for Mars with crust formation and mantle differentiation are compared. In the first model-termed the homogeneous differentiation model-we assume that a basaltic crust has grown steadily in 4.5 Ga as a consequence of pressure-release partial melting of mantle rock. The second model-termed the early differentiation model-incorporates the dichotomy and an early differentiation event. This event is assumed to have resulted in a mantle depleted of radioactive elements and a primordial enriched southern highland crust. We assume that the primordial crust acts as an efficient thermal blanket on the southerly hemisphere mantle. In a second stage of differentiation, a secondary basaltic crust in the northerly hemisphere is produced by pressure-release partial mantle melting. Our calculations suggest that the homogeneous differentiation model cannot explain the isotopic characteristics of the SNC-meteorites, the concentration of Ar-40 in the present Martian atmosphere, the dichotomy, and the long-term stability of the northerly hemisphere volcanism. The early differentiation model has the required geochemical reservoirs (depleted mantle, enriched crust and, possibly, subcrustal mantle layer) and the calculated volume of the secondary crust is consistent with the concentration of Ar-40 in the atmosphere. The thickness of the secondary crust is between 10 and 40 km. It depends mainly on the amount of mantle depletion and the crust production efficiency, but little on the amount of thermal blanketing of the mantle by the primordial crust. The lithosphere thicknesses in the two hemispheres, on the contrary, depend to a large extent on the amount of thermal blanketing and little on the other two of the above parameters. The present lithosphere thicknesses are roughly 150-200 km in the northerly hemisphere and 350-500 km in the southerly hemisphere. The present-day surface heat flow in the southerly hemisphere may be about 15 mW m-2 smaller than in the northerly hemisphere. Mantle temperature decreases with the amount of depletion of the mantle and increases with the amount of thermal blanketing, but differs by no more than about 100 K from mantle temperatures calculated in thermal evolution models that neglect differentiation. Therefore, earlier core evolution and magnetic field generation models [Stevenson et al., Icarus 54, 466-489 (1983); Schubert and Spohn, J. geophys. Res. 95, 14,095-14,104 (1990)] calculated without allowing for mantle differentiation remain essentially valid.
